NASA, CBP Tackle Data Access, Security Challenges Amid AI Expansion

NASA is working to make its internal data accessible through application programming interfaces (APIs), while Customs and Border Protection is expanding its data-sharing infrastructure to support growing AI capabilities, the agencies’ chief data officers said at GovCIO Media & Research’s Federal Cloud and Data Forum Thursday.

NASA acting Chief Data and Super Intelligence Officer Kevin Murphy said the effort requires coordination among NASA facilities and vendors when transferring large amounts of data. The agency must also ensure the process protects its proprietary information.

“You can’t have super intelligence or artificial intelligence without a really good data ecosystem. If you don’t have that, then you can’t do the things that we need to do at NASA,” Murphy said.

Murphy highlighted a recent agency directive that treats data as a strategic asset to accelerate super intelligence capabilities. The directive applies to NASA’s internal systems, including those supporting finance, engineering and research.

Big Data, Big Science

NASA has launched several pilot programs to improve data access and sharing in support of its research efforts.

“We’re trying to get back to the moon by 2028. We’re trying to launch better, faster science equipment. We’re trying to launch the next era of X planes. To do that, we really need to connect our wind tunnels to our computing environments, to the different databases of the engineering information that we have,” he said.

These efforts require moving data quickly and securely without navigating multiple authorization processes.

To achieve this, NASA aims to make internal information directly accessible to authorized users through its API ecosystem. The agency is modernizing its infrastructure to support increased data traffic and taking a similar approach to its publicly available scientific information.

“We have to be very careful about how we share that [information], and also how that information gets incorporated into any of the super intelligence models,” Murphy said.

The effort raises additional considerations, including how to efficiently train AI models using different types of data. Murphy explained that determining who can access information and how various data sources can be combined “can be a pretty tricky problem from a governance and policy perspective.”

He added that while NASA has identified approaches to some of these challenges, more work remains.

Data Sharing at CBP

Customs and Border Protection’s mission spans a wide range of responsibilities, from securing the nation’s borders and ports of entry to monitoring incoming trade goods to protect U.S. citizens and businesses, said Michelle Zebrowski, CBP’s CDO.

The agency has approximately 70,000 employees working across geographically dispersed operations, including air and marine missions, Border Patrol, the Office of Field Operations and the Office of Trade. These diverse responsibilities have resulted in a wide range of datasets across the organization.

Zebrowski noted that CBP also maintains several publicly accessible datasets. The agency’s size and operational diversity can make it difficult to share data across internal divisions or with the Department of Homeland Security. However, recent efforts to develop APIs have helped alleviate some of these challenges.

Although developing these capabilities takes time, Zebrowski said they establish a foundation for scaling data sharing across the agency.

Deep Experience with AI

CBP has extensive experience with AI, dating back approximately 15 years to its early work with machine learning systems, Zebrowski said.

The agency uses ChatCBP, a large language model-based tool operating within its internal network infrastructure. As CBP expands its data infrastructure, additional information will become accessible to support AI applications, she said.

Educating employees about these capabilities remains a priority. Zebrowski said her office spent a year working with different parts of the agency to emphasize the importance of cataloging datasets. Once employees understand the value of making data accessible, the technical process is relatively straightforward, she added.

“It’s just getting everybody on the same page with the importance of knowing what we have, so that we can govern it,” Zebrowski said.